[1N2-J-9] Natural language processing, information retrieval: dialogue

[1N2-J-9-01] Dialogue based recommender system that flexibly mixes utterances and recommendations

〇Daisuke Tsumita¹, Tomohiro Takagi¹ (1. Department of Computer Science, Meiji University)

Keywords:Dialogue System, Recommender System, Deep Reinforcement Learning

Many of the prior research in the recommendation through dialogue were designed separating dialogue and recommendation. However, since the accuracy of the recommendation itself is not necessarily high, rarely the recommendation result meets user needs. We human, however, can guide the solutions satisfying the user, by appropriately repeating the cycle of checking mismatch reason and making another recommendation in our conversations. In this paper, we proposed a system to leverage a dialogue strategy for reinforcement learning using recommendation results based on user’s utterances. We realized a dialog system to perform adaptive behavior that naturally incorporates recommendations into conversation with users.